Development and Validation of the Hospital Medicine Safety Sepsis Initiative Mortality Model
Authors' conclusion
The HMS-Sepsis mortality model showed strong discrimination and adequate calibration and reclassified one-third of hospitals to a different performance category from unadjusted mortality.
Study context
- Population
- Community-onset sepsis hospitalizations in the Michigan HMS-Sepsis registry, April 2022-September 2023
- Participants
- 17,514 hospitalizations · 66 hospitals
- Exposure
- Derivation vs Validation
- Derivationn = 12,260
- Validationn = 5,254
- Outcome
- Model performance for predicting 30-day mortality (C statistic, Brier score, calibration by deciles of risk)
- Follow-up
- 30 days
Methodological detailssplit derivation (70%) and validation (30%) cohorts, prediction model development and validation
- Split derivation (70%) and validation (30%) cohorts
- Prediction model development and validation
Main finding
As reported in the abstract
Discrimination and calibration of the HMS-Sepsis mortality model
| Group | C statistic |
|---|---|
| Derivation | 0.82 |
| Validation | 0.81 |
C statistic
Derivation0.82
Validation0.81
- C statistic for all subgroups assessed
- ≥ 0.78
- Overall calibration error
- 0.0%
- Mean calibration error across deciles of risk
- 1.5%
Observational data: associations, not effects.
Further findings
Hospitals with a different performance assessment using standardized mortality ratios vs observed mortality
- Hospitals reclassified
- 33.9%
30-day mortality, total cohort
- 30-day mortality
- 19.4%
Final model content
- Physiologic variables
- 13
- Physiologic interactions
- 2
- Demographic and chronic health variables
- 16
- Most significant variables: age, metastatic solid tumor, temperature, altered mental status, platelet count
Original publication
Development and Validation of the Hospital Medicine Safety Sepsis Initiative Mortality Model. Chest. Published 2 July 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →